Support NaN values in UpliftTree and UpliftRandomForest #860
Conversation
|
|
|
The Read the Docs build failure appears to be due to Cython extensions requiring a compiler in the RTD environment. |
jeongyoonlee
left a comment
jeongyoonlee
left a comment
There was a problem hiding this comment.
Thanks for your contribution, @aman-coder03. I left a few comments. Can you address them? Also, please add test code to tests/test_uplift_trees.py accordingly. Thanks!
| is_split_by_gt = False | ||
|
|
||
| for value in lsUnique: | ||
| len_X_l = group_counts_by_divide(columnValues, value, is_split_by_gt, treatment_idx, y, left_count_arr) |
There was a problem hiding this comment.
We need this code block. Is there any reason you removed it?
| n_reg | ||
| ) | ||
|
|
||
| early_stopping_flag = False |
There was a problem hiding this comment.
We need this code block for early stopping.
causalml/inference/tree/uplift.pyx
Outdated
There was a problem hiding this comment.
NaN handling is needed here in the percentile calculation.
|
@jeongyoonlee i will look into this! |
aman-coder03
force-pushed
the
feature/handle-nan-uplift-trees
branch
from
e13b20b to
bd23f63
Compare
3 participants
This PR adds native support for missing values (NaNs) in
UpliftTreeandUpliftRandomForest.Each candidate split evaluates both possible NaN routing directions
(left/right) and learns the optimal routing per node, similar to
scikit-learn’s decision tree behavior.
The learned NaN routing is stored in each tree node and applied
consistently during training, pruning, filling, and prediction.
This resolves #802.